INDEX

Explanations

names with a specific structure, containing "de" followed by a high-ranking second part

oai_token-act-pair · gpt-3.5-turbo

occurrences of the prefix "de" in names or terms

oai_token-act-pair · gpt-4o-mini Triggered by @bot

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 11-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.11.hook_resid_pre

Prompts (Dashboard)

24,576 prompts, 128 tokens each

Dataset (Dashboard)

Skylion007/openwebtext

Features

24,576

Data Type

torch.float32

Hook Point

blocks.11.hook_resid_pre

Architecture

standard

Context Size

128

Dataset

Skylion007/openwebtext

Hook Point Layer

Activation Function

relu

Embeds

PlotsExplanationShow Test FieldDefault Test Text

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

CPC

-0.84

phrine

-0.72

 amounts

-0.69

 inactive

-0.67

 exhibits

-0.65

 topical

-0.64

tons

-0.64

ilial

-0.63

 exhibit

-0.63

 institution

-0.63

POSITIVE LOGITS

Berry

1.14

 Blasio

1.09

Marco

1.05

Cola

1.00

Bo

0.99

Fran

0.97

Jesus

0.97

Leon

0.97

Gra

0.96

Shield

0.96

Activations Density 0.029%

names with a specific structure, containing "de" followed by a high-ranking second part

occurrences of the prefix "de" in names or terms

No Comments

No Known Activations